On Coexistence of Wireless Systems in Unlicensed Bands Using Cognitive Medium Access

We propose a holistic cognitive radio (CR) approach to tackle the coexistence problem of wireless systems in heterogeneous environments. Our approach designs and optimizes the cognitive capability of wireless systems by taking all tasks of a cognitive cycle into account: environment analysis, awareness, learning, and decision. The result is a medium access protocol that enables automatic coexistence in dense unlicensed bands. First, relying on a solid foundation of extensive measurements in targeted environments, we extract highly realistic traffic models for wireless channel occupancy, which are capable of reproducing key statistical properties in terms of self-similarity and repeatability. This constitutes the first building block of our approach, i.e. environment analysis and awareness. Next, we utilize the extracted models and their properties to realize the second task of the cognitive cycle, i.e. learning. To achieve this, we propose a novel spectrum prediction method based on Hidden Markov Models (HMM) that forecasts the exact behavior of the primary user (PU) for potentially very long periods of time, using the awareness achieved in the first task. The proposed algorithm exhibits high levels of accuracy for long-term prediction of channel activity. Next, we use the optimized prediction algorithm to realize the decision task of the cognitive cycle. Here, we integrate the learning and prediction cognitive capabilities in a medium access control (MAC) scheme and conceptualize spectrum access behavior of wireless systems, for both offline or online learning cases. We subsequently derive models for the medium access behavior in the presence of a preexisting PU on the channel. The proposed approach achieves an improvement in the overall reliability in terms of packet error rate by roughly two orders of magnitude for low number of nodes, and one order of magnitude for a high number of nodes compared to the traditional Carrier Sense Multiple Access (CSMA). Additionally, the proposed approach leads to an improvement in the protection of preexisting wireless systems by roughly one order of magnitude compared to CSMA for both high and low number of nodes, and an improvement in the normalized throughput by a factor of more than two for highly dense environments. The implications of those improvements are clear in terms of coexistence of wireless systems in challenging use cases such as Industry 4.0.